Time Variation in Asset Return Dependence: Strength or Structure?

This paper proposes a novel flexible approach to modelling time variation in asset return dependence by means of mixture copulas. We distinguish between the strength of dependence as determined by the parameter(s) of a given copula, and the structure of dependence as determined by the copula specification. The suggested time-varying mixture copula allows both dependence characteristics to vary over time. An empirical application to daily international equity markets, using a mixture of the symmetric, tail-independent Gaussian copula and the asymmetric, tail-dependent Gumbel copula, provides evidence that both the strength and structure of dependence exhibit substantial time-variation. From an economic perspective, we show that including changes in either the strength of dependence or the structure of dependence only, leads to serious miscalculation of Value-at-Risk estimates.